English

Large-Scale Gaussian Processes via Alternating Projection

Machine Learning 2024-03-12 v2 Machine Learning

Abstract

Training and inference in Gaussian processes (GPs) require solving linear systems with n×nn\times n kernel matrices. To address the prohibitive O(n3)\mathcal{O}(n^3) time complexity, recent work has employed fast iterative methods, like conjugate gradients (CG). However, as datasets increase in magnitude, the kernel matrices become increasingly ill-conditioned and still require O(n2)\mathcal{O}(n^2) space without partitioning. Thus, while CG increases the size of datasets GPs can be trained on, modern datasets reach scales beyond its applicability. In this work, we propose an iterative method which only accesses subblocks of the kernel matrix, effectively enabling mini-batching. Our algorithm, based on alternating projection, has O(n)\mathcal{O}(n) per-iteration time and space complexity, solving many of the practical challenges of scaling GPs to very large datasets. Theoretically, we prove the method enjoys linear convergence. Empirically, we demonstrate its fast convergence in practice and robustness to ill-conditioning. On large-scale benchmark datasets with up to four million data points, our approach accelerates GP training and inference by speed-up factors up to 27×27\times and 72×72 \times, respectively, compared to CG.

Keywords

Cite

@article{arxiv.2310.17137,
  title  = {Large-Scale Gaussian Processes via Alternating Projection},
  author = {Kaiwen Wu and Jonathan Wenger and Haydn Jones and Geoff Pleiss and Jacob R. Gardner},
  journal= {arXiv preprint arXiv:2310.17137},
  year   = {2024}
}

Comments

AISTATS 2024

R2 v1 2026-06-28T13:02:22.530Z